Slashing operational-costs-via-driveless-ran-optimization

White Paper

Slashing Operational Costs via Driveless RAN Optimization 1

Executive Summary

As mobile networks grow in size and complexity, efficient management of daily network

operations become one of the most crucial tasks of mobile network operators(MNOs). In this

white paper, we discuss how 3GPP defined Minimization of Drive Tests (MDT) data can be

utilized in lieu of drive testing to realize driveless Radio Access Network (RAN) optimization.

The approach outlined here does not require GNSS (Global Navigation Satellite System)

location information from mobile terminals or applications. Field results obtained by North

American Tier-1 mobile operators have shown that more than 50% reduction in operational

expenses is possible from drive test elimination when introducing new sites/clusters.

Introduction to MDT

LTE continues to grow faster than any other mobile communications system technology in

history since its introduction in December of 2009. As of January 2016, 480 operators have

commercially launched LTE systems, reaching to 13% of mobile connections worldwide.

Similar to 3G, LTE deployments introduce significant operational and capital expenditures for

mobile operators. Traditionally, test terminals are used to measure signal and service quality

levels (e.g. RSRP, RSRQ, SINR, throughput, access fails and drops) during LTE deployments.

Drive test logs are collected and analyzed to rectify any issues encountered. This is a very

tedious process where large log files need to be handled and analyzed with post processing

tools to come up with conclusions. In practice, the whole drive testing process has to be

repeated several times due to equipment misconfigurations, failures noticed from earlier tests

and important routes which were not driven previously. Also, after each time RF related

optimization changes are applied (e.g. electrical antenna tilt changes), drive testing analysis is

typically re-run to assess the new results from the field.

3GPP, a global partnership that defines the rules of today’s mobile communications systems,

specified the standards of how LTE networks should produce data to minimize traditional drive

testing efforts under the Minimization of Drive Tests (MDT) initiative.

In this white paper, we discuss how 3GPP defined MDT data can be used to implement a

“driveless” RAN optimization framework for mobile operators to dramatically reduce their

operational expenses.


Shortcomings of Drive Testing

Since the 1990s, mobile subscribers have experienced a dramatic increase in network

capacity, starting with 9.6 Kbps CS download rates in GSM to more than 300 Mbps with LTE-

A. However, RAN optimization processes to `tune` mobile networks have not changed much

within this period to keep up with the advancements.

In a typical LTE network deployment scenario, the first step is to design the network while

taking into account the expected subscriber base, service quality committed to end users, and

the RAN budgets, the most expensive part of CapEx. Then, eNBs are deployed as designed

and budgeted typically with lower density over an existing 2G/3G radio network layer. At this

`pre-launch` phase, drive tests are performed to simulate subscriber behavior and to identify

any immediate issues with the deployment. Following the commercial launch of planned sites,

more eNBs are gradually added to enhance both the coverage and the capacity of the network

while introducing minimum disturbance for commercial users. Unless the parameters are

properly tuned, new site additions can potentially disturb existing neighborhood sites. Because

of the need to test the footprint of each added site’s coverage, drive testing becomes an

important task in this phase.

Shortcomings of drive testing during this `post-launch` phase are as follows:

Long testing duration required: Assuming an urban region with average site-to-site distance of

1-3 miles where 200 eNBs are on-aired, every new eNB added to the region after cluster

launch requires an average of 10-15 miles of drive testing (excluding drives to get/return to the

site). This corresponds to 1,000 – 1,500 miles of driving if 100 eNBs are added. This

significantly increases the time it takes the MNOs to realize revenues from the new assets

added.

High volumes of data to process: If we consider that Tier-1 operators have thousands of eNBs,

the amount of drive test data that is required to be collected and analyzed amounts to huge

volumes. For a 25-eNB urban cluster in a 3 miles x 1 miles area with 4 pieces of test equipment

including a scanner searching three separate bands, the amount of drive data can exceed 1

GB of raw files for a 7-hour drive.

Sampling limitation: It may take a large number of tests to be repeated in order to replicate a

specific drop or a block problem. More tests mean more logs and consequently result in wasted

resources.


Indoor user experience: Drive testing is performed in outdoor environments. However, the

majority of subscriber traffic takes place indoors. Thus, conventional drive testing approach

cannot verify and improve indoor subscriber experience.

MDT as a Driveless Solution for Post-Launch Optimization

As a response to the shortcomings of drive testing, 3GPP published Minimization of Drive

Tests (MDT) specifications [1, 2, and 3] to provide a more efficient approach to optimization.

Using the measurements taken from MDT-supported equipment, operators can select and

display all or a portion of the UEs (User Equipment) under specified eNBs covering a particular

geographical region or specific IMSIs, IMEIs or IMEI-TACs [4] across the network. These

standardized UE and equipment measurements are then used for various needs including new

site/cluster RF Tuning or VIP customer complaint handling. The output measurements of MDT

are also normalized to be used as inputs to SON (Self Organizing Networks) use cases and

algorithms such as CCO (Coverage and Capacity Optimization) [5, 6].

Driveless MDT optimization complements the mobile operator`s `post-launch` deployment

process as follows:

Step 1. Site installation and eNB integration to OSS

Step 2. Identification and resolution of hardware issues (e.g. PIM, RSSI, VSWR)

Step 3. eNB and MME configuration audits (e.g. golden parameters, TAC, PCI, RSI)

Step 4. Pre-launch single site audit (e.g. crossed feeder/MIMO verification,

stationary DL/UL throughput check, testing of Tx imbalance issues, co-site handover

check)

Step 5. eNB and cluster launch

Step 6. Post-launch optimization with driveless MDT data

Once an eNB or cluster is launched for commercial traffic, driveless RAN optimization process

starts. MDT configuration is done by specifying eNB/E-UTRAN cell list, time, duration of

measurement, messages, events to be collected over Uu and eNB external interfaces and

sampling rate of calls (e.g. 50%). Typically, files are created per 15-minute intervals on eNBs

and then forwarded to OSS (Logged MDT). Streaming transfer option directly to an external

server is also defined in the standards (Immediate MDT).


KPI results are calculated and analyzed per geographic bin where resolution depends on

operator requirements (e.g. ranging from 50m x 50m to 500m x 500m).

Various Key Performance Indicators (i.e. KPIs1) ranging from availability, accessibility,

retainability, mobility, integrity and throughput for different services (e.g. PS Data, VoLTE,

CSFB) should be above target performance thresholds set by the mobile operator for best

subscriber experience.

An important benefit of the driveless approach is that, the operator can improve service quality

experienced by subscribers in every part of the network per geographic bin. Driveless post-

launch optimization improves bad subscriber experience even for cells with very good network

KPIs, which reflect the accumulated results of all subscribers in a cell.

Additionally, geographic bin analysis facilitates policy tuning per bin, where important KPIs

(e.g. throughput and coverage) are weighted with larger coefficients than others during the

calculation of a unified quality index per bin.

Figure 1 - Cost calculation example for a geographic bin

1 3GPP initial output to extend traditional KPI definition to Service Experience KQI is described under

TR 32.862. [7,8]


The following Figure 2 - Drive Testing vs. Driveless Tuning of a newly added site/cluster to

commercial network compares Drive Testing with driveless MDT based network tuning to

emphasize the shortened duration and minimized effort of the driveless solution.

Figure 2 - Drive Testing vs. Driveless Tuning of a newly added site/cluster to commercial network


Comparison of MDT Alternatives

Some alternative solutions to MDT with their pros and cons are provided below:

Probe systems: This solution requires external equipment to be deployed in the RAN and the

Core Network. It is useful for end-to-end verification of applications and troubleshooting, but

lacks all other requirements of a new site and cluster `post-launch`. The cost is high.

Geolocation tools: Typically, these solutions focus on visualization of events such as drops,

blocks, throughput and radio coverage for reporting and troubleshooting purposes. They lack

the capability to automatically update RF parameters needed to bring improvements to the

network.

SON tools without geolocation support: SON solutions that do not support geolocation have

the capability to provide some RF tuning recommendations such as dynamically tuning mobility

thresholds, offsets or hysteresis. These tools lack cell footprint visualization and optimization

capabilities as required by new site and cluster `post-launch` processes.

Support for

Traditional

Drive Test

tools

Probe

systems

Geolocation

tools

SON tools

w/o

Geolocation

support

Driveless

MDT

solution

New site & Cluster RF Tuning

(visualization of drop, block

events and coverage on map)

Yes No Yes No / Limited Yes

Provide RF tuning

recommendations No No No Yes Yes

Compatibility with other SON

functions such as CCO No No / Limited No / Limited No Yes

Verify indoor customer

experience No No Yes No Yes

VIP customer complaint

handling No / Limited No / Limited Yes No Yes

End to end verification of

applications and

troubleshooting

Yes Yes No No No

Competitor benchmarking Yes No No No No

Table 1 - Comparison of Driveless MDT solution with other alternatives


Case study: Downlink throughput verification and tuning

using driveless optimization

Following figures depict driveless MDT data analysis results of a cluster during post-launch

optimization. Figure 3 shows the first carrier cells in low band (800 MHz) whereas Figure 4

shows the second carrier cells in the high band (1800 MHz) of the same eNBs in the cluster.

Subscribers’ downlink throughput values per bin are optimized during post-launch, which takes

into account configured band priorities. (i.e. for the high priority 1800 MHz band, subscribers

shall be served close to site, whereas 800 MHz band subscribers shall be served mostly at cell

borders)

Figure 3 - User downlink throughput (purple: 5Mbps to 10Mbps) on base 800 MHz band LTE carrier cells

As shown in Figure 2, subscribers served at cell edge areas (either outdoor or indoor) on the

800 MHz band experience 5 to 10 Mbps throughput. Other subscribers that are close to cell

center areas enjoy downlink throughput rates higher than 30 Mbps on the1800 MHz band. With

this band separation, user location and user throughput, the operator can control the quality of

experience of subscribers as a result of detailed band, location and throughput view in the post

launch cluster. In accordance with operator policies, recommendations in the form of electrical

antenna tilt and power changes are applied to improve subscriber experience.


Figure 4 - User downlink throughput (green: >= 30Mbps) on overlay 1800 MHz band LTE carrier cells

Drive test throughput measurements are also correlated on the same map using the same

colored legend. Due to the inherent mobility of drive tests, downlink throughput testing results

obtained around the same locations are worse than the throughput observed by MDT UEs,

which are mostly stationary.

These observations provide valuable insight as to how traffic and throughput is dispersed over

the testing region. Further, it also helps to verify the layer management strategies of the

operators. In this specific example, subscribers on high band carrier that are served close to

site experience greater than 30 Mbps throughput whereas subscribers at cell borders get

around 8 Mbps throughput on low band carrier. Without detailed breakdown of UE provided

data, overall computed KPIs would not provide the details on how the throughput is divided

among subscribers. Improvements can be fine-tuned based on MDT data obtained from UEs

in various periods and durations.


Conclusion

Driveless optimization based on MDT data is an efficient way to handle the performance tuning

and verification of today’s mobile networks of growing complexity, compared to traditional drive

test based methods. Using MDT, the cost and the duration of post-launch optimization process

is greatly reduced. This provides faster roll out times with optimum user experience.

MDT data provides the framework where hidden problems in the network are revealed and

effective solutions are implemented in a quick and cost efficient way. This approach

accumulates measurements from all subscribers throughput the network as opposed to a drive

testing performed with limited number of test UEs on limited regions.

P.I. Works has extensive expertise in post-launch optimization using MDT data with various

operators. For more information, please contact P.I. Works for Driveless RAN Optimization

Solutions.

About P.I. Works

P.I. Works, is a leading provider of next-generation Radio Access Network (RAN) management

solutions. P.I. Works’ expertise in mobile network optimization, which spans over a decade,

combined with the commercially available product portfolio and services, enables global Mobile

Network Operators (MNOs) to improve network quality and subscriber experience, while

increasing profitability.

To date, P.I. Works has deployed its solutions for 38 MNOs in 27 countries.

P.I. Works state-of-the art product portfolio, unified Self Organizing Networks (uSONTM),

automates the optimization and operational tasks of complex mobile networks 24/7 to increase

quality, capacity and coverage.

For more information, please visit http://www.piworks.net/ or send e-mail to [email protected]

Copyright© 2016 P.I. Works, All Rights Reserved.

The information contained in this document is the property of P.I. Works. Except as specifically authorized in writing by P.I. Works, the holder of this document shall keep the information contained herein confidential and shall protect same in whole or in part from disclosure and dissemination to third parties and use same for evaluation, operation and maintenance purposes only. The content of this document is provided for information purposes only and is subject to modification. It does not constitute any representation or warranty from P.I. Works, as to the content or accuracy of the information contained herein, including but not limited to the suitability and performances of the product or its intended application. uSON and the P.I. Works logo are trademarks of P.I. Works. All other trademarks are the property of their owners.

mailto:[email protected]

http://www.piworks.net/

mailto:[email protected]

https://www.linkedin.com/company/piworks-inc-


References

[1] 3GPP TS 32.421: Subscriber and equipment trace: Trace concepts and requirements

[2] 3GPP TS 32.422: Subscriber and equipment trace: Trace control and configuration

management

[3] 3GPP TS 32.423: Subscriber and equipment trace: Trace data definition and management

[4] S5-112666 TS 32.422: Enhancement for MDT Initiation with IMEI-TAC usage [P.I. Works

contribution to 3GPP, Aug 2011]

[5] S5-154288 TS 28.628 Add NM-Centralized CCO related measurements and delete

redundant ones [P.I. Works contribution to 3GPP, Aug 2015]

[6] S5-154289 TS 28.628 Correct coverage hole definition for NM centralized Coverage and

Capacity Optimization (CCO) [P.I. Works contribution to 3GPP, Aug 2015]

[7] 3GPP TR 32.862: Study on Key Quality Indicators (KQIs) for service experience

[8] S5-153343 TR 32.862 pCR 32.862 NGMN Service Quality Definition and Measurement

for HTTP Adaptive Streaming [P.I. Works contribution to 3GPP, May 2015]

Slashing operational-costs-via-driveless-ran-optimization

Services